03 · What You Need to Know
Institutions Can Be Part of the Mechanism, Not Merely the Background
Research findings are produced inside systems
People do not encounter interventions, policies, technologies, or services in an institutional vacuum. Healthcare is delivered through financing and referral systems. Education operates through curricula, assessment rules, teacher preparation, schedules, governance, and resource allocation. Policies interact with enforcement structures, incentives, administrative capacity, and existing regulations.
These arrangements can affect who receives an intervention, how it is delivered, whether people participate, what alternatives are available, and what consequences follow.
The updated Medical Research Council guidance for complex interventions explicitly treats context as dynamic and multidimensional, including organizational, social, cultural, political, economic, physical, and spatial features. Intervention effects can depend on these conditions.
Institutional context can therefore be scientifically substantive rather than a paragraph of background information.
Institutional difference
Two settings have different systems, policies, organizational structures, or procedures.
Scientifically relevant institutional difference
A particular structural feature changes a mechanism, exposure, incentive, implementation condition, or outcome relevant to the research claim.
Ask where the institution enters the causal pathway
Suppose research shows that sending appointment reminders increases attendance at medical consultations. Would another healthcare system require a new effectiveness study?
Not necessarily. But imagine that the original studies occurred where appointments are centrally scheduled, patients have stable contact information, transportation is accessible, and consultation costs are largely covered. In the target system, missed appointments may arise primarily from substantial out-of-pocket costs or inability to obtain transportation.
The reminder itself may work perfectly as a reminder while doing little to remove the dominant barrier to attendance.
The relevant difference is not simply “another healthcare system.” It is that the system changes the constraints connecting the intervention to the desired outcome.
Healthcare systems can alter access, incentives, and implementation
Healthcare findings may depend on insurance coverage, out-of-pocket costs, referral pathways, workforce availability, medicine supply, service location, professional scope of practice, reimbursement, continuity of care, and many other system characteristics.
An intervention evaluated in a setting with abundant specialists may not be implementable through the same pathway where specialists are scarce. A screening program may have different consequences when positive cases can be referred promptly than when diagnostic or treatment capacity is limited.
This does not mean every healthcare intervention requires a trial in every health system. It means that applicability should be assessed in relation to the features required for the intervention to produce its expected benefit.
Education systems can alter what the same intervention actually becomes
Educational interventions are similarly embedded in systems. Curriculum requirements, class size, assessment practices, instructional time, teacher autonomy, professional development, technology access, language of instruction, and accountability structures can all affect implementation.
Consider an instructional intervention requiring teachers to reorganize lesson sequences in response to formative assessment. Evidence from systems where teachers have substantial curricular discretion may not transfer straightforwardly to a system with tightly prescribed pacing and assessment schedules.
The educational mechanism may be sound, but the institution constrains whether teachers can activate it.
A useful local study would therefore investigate the institutional constraint and its consequences rather than simply repeating the intervention and attributing any discrepancy vaguely to “context.”
Policy effects depend on the surrounding policy environment
A policy does not operate independently of other rules. Its effects may depend on enforcement, administrative capacity, incentives, sanctions, exemptions, public awareness, existing legislation, political legitimacy, and available alternatives.
A regulation that changes behavior in one jurisdiction may produce a different effect where enforcement is weaker or where regulated actors can easily shift to an unregulated substitute.
Policy research therefore requires careful specification of the treatment itself and the institutional environment in which it operates. A policy with the same formal name may function differently across jurisdictions because implementation differs.
Organizations can alter behavior through incentives and routines
Institutions influence individual behavior partly by shaping incentives, opportunities, expectations, information, and routines.
A workplace intervention might depend on supervisors providing employees with protected time. A university policy might depend on instructors applying assessment rules consistently. A hospital protocol may depend on communication between departments.
If the target institution organizes these processes differently, another study may provide useful evidence. But again, the institution's identity is not the explanatory variable. The relevant organizational process is.
Implementation research is particularly useful when effectiveness is established but delivery is uncertain
One of the most important consequences of institutional difference is that researchers may need a different kind of study rather than another efficacy trial.
WHO defines implementation research as the scientific study of processes used to implement policies and interventions and the contextual factors affecting those processes. It focuses on how evidence-based interventions function in real-world settings.
If an intervention already has strong evidence of efficacy but local institutional arrangements create uncertainty about access, acceptability, reach, fidelity, adoption, or sustainability, implementation research may be more useful than asking from the beginning whether the intervention can work at all.
WHO's ethics guidance also makes an important qualification: local implementation research should not be assumed necessary merely because the location is new. There should be reasonable doubt that contextual differences could affect implementation.
Institutional differences can change implementation without changing the intervention on paper
Two sites may formally adopt the same program while delivering substantially different interventions in practice.
One healthcare system may provide the intervention through specialists, another through primary-care staff. One school may allocate weekly instructional time, another only occasional sessions. One agency may enforce a policy consistently, another sporadically.
Implementation outcomes such as fidelity, reach, adoption, acceptability, feasibility, and sustainability can therefore mediate the relationship between formal intervention and observed outcome.
When these pathways are plausible, measuring implementation becomes important for interpreting cross-context differences.
Institutional context can change absolute consequences even if relative effects are similar
A finding can transfer scientifically while producing different practical consequences.
Suppose an intervention reduces a particular risk proportionally across settings. If baseline risk differs because of institutional conditions, the absolute number of outcomes prevented may differ substantially. Costs and opportunity costs may also vary.
Decision-makers therefore sometimes need local institutional data even when the relative effect itself is not seriously in doubt.
This helps explain why local evidence can sometimes be especially useful for a local decision without implying that stronger international evidence should be discarded.
Do not confuse institutional context with resources
Institutions and resources overlap, but they are not identical.
A hospital may possess sufficient financial resources but organize referrals inefficiently. A school system may have adequate devices but policies that prevent students from using them in the relevant activities. Conversely, an institution may have excellent procedures but insufficient personnel to implement them.
If the central uncertainty concerns staffing, infrastructure, funding, time, equipment, or other capacity constraints, the neighboring question of whether resource differences justify another study may provide the more precise framing.
Institutional labels can hide substantial variation within systems
Researchers often compare “public versus private,” “urban versus rural,” “centralized versus decentralized,” or one national system against another. These categories can be useful descriptions, but they may conceal considerable variation.
Two public hospitals may differ in staffing, leadership, patient population, referral networks, implementation climate, and available services. Schools governed by the same national curriculum may implement it differently. Local enforcement of a national policy may vary considerably.
Researchers should therefore measure relevant institutional characteristics whenever possible rather than assuming that a broad institutional category explains the outcome.
Existing evidence may already span the institutional variation you care about
Before proposing another study, examine whether previous research has already tested the phenomenon across diverse systems.
If an intervention has produced similar results under multiple financing arrangements, organizational structures, resource levels, and implementation models, another institution that falls within this range may add relatively little.
If existing evidence comes almost entirely from one institutional model, a meaningfully different system may provide a much stronger test.
This is part of determining whether the local context is different enough to warrant another study.
Do not infer an institutional effect from two different study results
Suppose a study in one healthcare system reports a large effect and your study reports a smaller one. The difference does not automatically demonstrate that the healthcare systems caused the discrepancy.
The studies may differ in populations, measurement, implementation, sample size, analytical decisions, or risk of bias. Direct comparisons, appropriate interaction analyses, multisite designs, or cumulative evidence may be needed to investigate institutional effect modification convincingly.
Watch Out
“The systems are different” is not an explanation for conflicting results. Identify which institutional feature differs, show why it could influence the mechanism or implementation, and collect evidence capable of evaluating that explanation.
The strongest institutional studies explain what feature could matter elsewhere
A study becomes more useful beyond its immediate setting when researchers describe institutional characteristics precisely.
“This intervention worked in Hospital X” tells readers relatively little about whether it will work in Hospital Y. “The intervention remained effective when delivered by generalist staff through an existing primary-care referral pathway with limited specialist support” provides information that other systems can compare with their own conditions.
This is one way a local study can contribute beyond its immediate setting. The institution becomes evidence about a type of condition rather than merely a named site.
